Real-Time Foot Clearance Biofeedback to Assist Gait Rehabilitation Following Stroke: a Randomised Controlled Trial Protocol

Background The risk of falling is signicantly higher in people with chronic stroke and it is, therefore, important to design interventions to improve mobility and decrease falls risk. Minimum Toe Clearance (MTC) is the key gait cycle event for predicting tripping-falls because it occurs mid-swing during the walking cycle where forward velocity of the foot is maximum. High forward velocity coupled with low MTC increases the probability of unanticipated foot-ground contacts. Training procedures to increase toe-ground clearance (MTC) have potential, therefore, as a falls prevention intervention. The aim of this project is to determine whether augmented sensory information via real-time visual biofeedback during gait training can increase MTC. Methods Participants will be over 18 years, have sustained a single stroke (ischaemic or hemorrhagic) at least 6 months previously, able to walk 50 metres independently and capable of informed consent. Using a secure web-based application (REDCap) 150 participants will be randomly assigned to either no-feedback (Control) or feedback (Experimental) groups, all will receive 10 sessions of treadmill training for up to 10 minutes at a self-selected speed over ve to six weeks. The intervention group will receive real-time, visual biofeedback of MTC during training and will be asked to modify their gait pattern to match a required “target” criterion. Biofeedback is continuous for the rst six sessions then progressively reduced (faded) across the remaining four sessions. Control participants will walk on the treadmill without biofeedback. Gait assessments are conducted at baseline, immediately following the nal training session and then during follow-up, at 1, 3 and 6 months. The primary outcome measure is MTC. Monthly falls calendars will also be collected for 12 months from enrolment. Discussion The project will contribute to understanding how stroke-related changes to sensory and motor processes inuence gait biomechanics and associated tripping risk. The research ndings will guide our work in gait rehabilitation following stroke and may reduce falls rates. Treadmill training procedures incorporating continuous real-time feedback may need to be modied to accommodate stroke patients who have greater diculties with treadmill walking. This innovative study will evaluate the impact of augmented sensory information for improving gait function, specically foot-ground clearance, via visually presented biofeedback. The research ndings will contribute to the broader falls prevention initiative by demonstrating the effectiveness of toe clearance biofeedback in making walking safer. It will also conrm whether faded feedback enhances learning by demonstrating how performance is affected immediately following training and whether it can be retained in the long term. If shown to be effective, there will be opportunities for biofeedback-training applications to other gait-impaired clinical populations. The project’s ndings will contribute to understanding how stroke-related changes to sensory and motor processes inuence the tripping risks of everyday walking. The protocol is, however, limited to people with chronic stroke capable of walking on a treadmill due to the continuous real-time feedback used for gait training.

ground clearance (MTC) have potential, therefore, as a falls prevention intervention. The aim of this project is to determine whether augmented sensory information via real-time visual biofeedback during gait training can increase MTC. Methods Participants will be over 18 years, have sustained a single stroke (ischaemic or hemorrhagic) at least 6 months previously, able to walk 50 metres independently and capable of informed consent. Using a secure web-based application (REDCap) 150 participants will be randomly assigned to either no-feedback (Control) or feedback (Experimental) groups, all will receive 10 sessions of treadmill training for up to 10 minutes at a self-selected speed over ve to six weeks. The intervention group will receive real-time, visual biofeedback of MTC during training and will be asked to modify their gait pattern to match a required "target" criterion. Biofeedback is continuous for the rst six sessions then progressively reduced (faded) across the remaining four sessions. Control participants will walk on the treadmill without biofeedback. Gait assessments are conducted at baseline, immediately following the nal training session and then during follow-up, at 1, 3 and 6 months. The primary outcome measure is MTC. Monthly falls calendars will also be collected for 12 months from enrolment. Discussion The project will contribute to understanding how stroke-related changes to sensory and motor processes in uence gait biomechanics and associated tripping risk. The research ndings will guide our work in gait rehabilitation following stroke and may reduce falls rates. Treadmill training procedures incorporating continuous real-time feedback may need to be modi ed to accommodate stroke patients who have greater di culties with treadmill walking.

Background
Stroke affects more than 60,000 Australians every year with 50% unable to walk one week following the event1. Impaired walking impacts independence by reducing the ability to perform everyday activities and limiting community participation2, 3. Falls risk is signi cantly higher in people with chronic stroke4 and approximately 50% of people living at home after a stroke will fall within 12-months11, with up to half sustaining multiple falls. Furthermore, in community dwelling people with stroke, up to 77% of falls occurred during walking. While there has been considerable research investigating falls risk management for older people generally, high-risk groups such as those who have had a stroke have not been extensively studied with respect to targeted falls prevention. Traditional exercise-based falls prevention programs are useful for the general older adult community but are not effective in people with stroke. For example, Batchelor et al. found that a multifactorial intervention including a home-based balance and strength program did not reduce falls in people with stroke 12. Another study con rmed that a group and home-based exercise program incorporating balance and strength training did not reduce falls 13. This suggests that alternative, targeted treatments to reduce falls risk in people with stroke are urgently needed.
Stroke adversely affects sensorimotor function and muscle strength, inhibiting the capacity to activate appropriate muscles and increasing the risk of contact between the foot and either the supporting surface or objects on it. Said et al. 21 found, for example, that stroke participants who had di culty in stepping over small obstacles (4-cm high) had greater falls rates. The key gait variable for predicting tripping-falls is Minimum Toe Clearance (MTC), an event mid-swing in the walking cycle6-9. Low MTC increases the probability of unanticipated foot-ground contacts8. Given that tripping directly results from unsuccessful toe clearance, previous research with both young and older populations has focussed on toe trajectory control during walking7-9, 17-19. Training individuals to increase MTC, therefore, has potential as a falls prevention intervention.
The aim of this project is to determine whether real-time biofeedback of toe clearance during gait training can signi cantly minimize tripping risk in people with stroke. We will test the e cacy of real-time biofeedback as an intervention to increase MTC using a RCT design incorporating both a training or "acquisition" phase with biofeedback. Retention tests will be conducted to con rm learning, as demonstrated by the longer-term or "relative permanence" of the targeted behaviour.
The primary objective is to determine whether real-time biofeedback of MTC during gait training will signi cantly increase MTC in people with stroke. We will also determine whether changes in MTC achieved on a treadmill transfer to overground walking. It is hypothesised that, compared to nobiofeedback training, visual biofeedback of foot clearance parameters during gait training will signi cantly increase toe-ground clearance (MTC), and MTC during biofeedback training will be retained in the longer term. It is also hypothesized that increases in MTC demonstrated in treadmill training will transfer to overground walking, such that tripping-risk in people with stroke is signi cantly reduced.

Methods
This single-blinded parallel group randomised controlled trial (RCT) with 1: 1 randomisation will assess the effects of biofeedback on MTC following gait training. It will conform to CONSORT guidelines and has been registered on the Australian and New Zealand Clinical Trials Registry (ANZCTR): ACTRN12617000250336. Ethics approval for the study was obtained from Austin Health, Melbourne Health and Victoria University Human Research Ethics Committees. Informed consent procedures include the provision that participants clearly understand that they are free to withdraw at any time without providing a reason. The study design owchart is shown in Figure 1. Eligibility criteria for participants Participants will be over 18 years, have sustained a single stroke (ischaemic or hemorrhagic) at least 6 months previously, able to walk 50 metres independently with or without a single point stick and capable of providing informed consent. Exclusion criteria are: (i) an ankle foot orthosis, (ii) neurological, orthopaedic, cardiac, respiratory or other medical conditions in addition to stroke that impact their ability to walk on a treadmill, (iii) over 158 kg body mass (due to the weight limit of the harness) and (iv) visual problems or severe visual-spatial neglect. Participants will not be receiving physiotherapy for their walking or lower limbs while enrolled in the study.

Participant recruitment study settings
Recruitment sites include Heidelberg Repatriation Hospital (Austin Health), the Royal Melbourne Hospital (Royal Park Campus) and Western Health. Interested parties can also self-refer in response to advertising.
Gait assessments and training are conducted at the Heidelberg Repatriation Hospital (Department of Physiotherapy) and the Victoria University (Footscray Park) Biomechanics Laboratory.

Randomisation
Randomisation to participant group will be performed by the project coordinator using a secure, webbased Research Electronic Data Capture (REDCap) tool hosted at Victoria University50. The participant group allocation will then be e-mailed to the intervention therapist with the subject heading Randomisation of (Participant ID). Participants will be coded and only available to authorized project personnel.
Blinding Clinical Physiotherapists will be involved in assessment but blinded to group allocation and training sessions. Blinded Clinical Physiotherapists will conduct clinical assessments and collect falls-related data. Biomechanical gait assessors who conduct the movement analysis are, similarly, blinded to group allocation to ensure that all assessment personnel are blinded as to the participant's group assignment.
Physiotherapists and Biomechanists delivering the gait training are required to know the patient's group assignment.

Intervention
All participants, both controls and those in the feedback intervention group, will walk on the treadmill following an identical schedule of training sessions. Self-selected walking speed will be determined following treadmill familiarization with a Physiotherapist during the rst clinical assessment visit. This self-selected walking speed will be recorded and each treadmill walking session will be conducted at the same walking speed. All participants wear a safety harness when treadmill walking and continue for up to 10 minutes with rest breaks as required. Previous research by the authors had determined 10 minutes to be comfortable maximum for this population29, 30. Participants will wear the same comfortable shoes for all walking activities.
Participants randomized to the intervention group undertake biofeedback gait-training using a real-time display of the affected limb's swing phase trajectory, with toe clearance and the associated MTC event clearly shown. Motion analysis markers will be attached to the shoe and other body segments to calculate toe clearance parameters in real-time as per the established protocol7. Target MTC will be calculated as baseline MTC plus 1 SD. The feedback group will be asked to move their affected limb such that MTC falls between target MTC mean+0.5SD and target MTC-0.5SD, projected as parallel lines on a screen in front of the treadmill35, 36. If baseline MTC cannot be calculated, target boundaries will be set using the maximum Toe Clearance (TC) with instructions to control foot motion within that target band. Borg Ratings of Perceived Exertion will be recorded after each session.
Biofeedback is presented for the rst six sessions then progressively reduced (faded) across the remaining four sessions. During fading one third of the initial feedback will be available at the beginning of session seven, one third in the middle for session eight, one third at the end of session nine, and nally one tenth at the beginning and one tenth at the end of session ten. As indicated above the control group will walk under the same conditions without either feedback or any instructions concerning their gait control. Adherence to the gait training program will be re ected in the number of training sessions successfully completed, and the research team will monitor any non-adherence. · Tardieu scale42, to assess spasticity in gastrocnemius and soleus; · Six minute walk test43, a measure of walking endurance; · Falls Risk for Older People in the Community (FROP-Com). The FROP-Com has previously been used to identify falls risk in people with stroke12.
Gait assessment trials are conducted at baseline, after the nal training session (a minimum of 20 minutes following nal training) and one-, three-and six-months later. These gait data are collected during treadmill and overground walking by biomechanics personnel blinded to group allocation. During overground walking assessments (pre and post training) participants walk at preferred speed along an 8m walkway. Kinematic (position/time) data are captured using a three dimensional motion analysis system (Optotrak®, NDI, Canada) with clusters of infrared markers xed to the toe of the shoe, and the shank, thigh and pelvis segments. Bony landmarks on the foot, ankle, knee and hip are identi ed using virtual markers and used with the clusters in order to construct a model of the joints and segments. This enables kinematic data, including joint angles, to be calculated. Foot plantar pressures and centre of pressure excursion (COPE) are captured to assess gait stability using the Pedar foot pressure insole (Novel, Germany), an in-shoe system that measures plantar pressure distribution within the shoe.

Primary outcomes
The primary outcome variables from the motion analysis, MTC magnitude and variability, will be assessed using biomechanical assessment data at baseline and post-training, which occurs immediately following the nal training session after a minimum of 20 minutes rest. Retention MTC data at 1, 3 and 6 months for the affected limb will determine whether the target MTC, ie baseline MTC plus 1 SD, has been maintained after visual biofeedback training has concluded. The precise probability of tripping due to foot-ground contact will be computed by statistically modelling the MTC histogram8.

Secondary outcomes
To supplement the MTC data, gait kinematics (secondary outcome variables) will be used to assess gait training effects on gait speed, stance/swing times, and joint angles at key events, foot-contact, toe-off and MTC of the affected and unaffected limb. Baseline and retention clinical data will show how falls risk measures were in uenced by feedback training. The association between the clinical tests and the biomechanical variables will be shown by correlating tripping probability and clinical measures. Mediolateral and anterior-posterior foot pressure (COPE) shifts will reveal biofeedback effects on gait stability. A further secondary outcome is the number of falls following training. These data will be collected via monthly falls calendars for 12 months from enrolment in the study as illustrated in the SPIRIT studytimeline ( Figure 2). In determining sample sizes, for the between-group comparisons, to detect differences in MTC a medium effect size (ES=f=0.25) between the group effects (No-Feedback vs Feedback) across the two time points (Baseline and post-training) a total of 126 or 63 per group is required (G*Power 47 calculation, 2x2 repeated measures design, α=0.05, power =0.85, correlation=0.7). To allow 20% drop out we shall initially recruit 75 participants per group (N=150). This sample size will also allow detection of a 1SD increase in MTC of the biofeedback group, compared to their pre-training. The probability of tripping8 calculated using the pilot (stroke) participants' data suggests that a 1 SD (0.77cm) increase in MTC will result in signi cant reduction in the risk of tripping on small (1 -2 cm) obstacles. A 0.77 cm increase in MTC represents approximately a 50% gain in minimum ground clearance (mean MTC = 1.4 cm) that is shown by our modelling to represent a substantial decrease in tripping risk, i.e., the probability of contacting an obstruction of a given height.

Data management and statistical analysis
All data will be stored via REDCap, an application speci cally designed for the safe storage of clinical research data, for 7 years. Only research personnel associated with the study will have access to the data and it will not be available to external agencies. Optotrak, IMU and Pedar data will be coded so no personal information will be identi ed in the data. De-identi ed data will also be stored on a secure drive at the University. The patient's gait kinematics are monitored as the project progresses to con rm no adverse effects of gait training, such as increased tripping risk. External auditing is not required in this study because all experimenters are trained to identify and record adverse events.
MTC data obtained from the treadmill walking will be analysed using two intervention group comparisons and repeated measures ANOVA of the intervention groups and change scores, with post-hoc comparisons where relevant. Walking speed will be used as a co-variate due to its in uence on MTC data. The between subject factor will be treatment group (No-feedback and Feedback) with Time (pre-training baseline, post-training, 1-month, 3-month and 6-month) being the within-subject factor. 1) Between-group comparisons of MTC post-training will reveal effectiveness of the biofeedback gait training method in improving MTC.
2) The within-subject analysis of the Feedback group will reveal the biofeedback effects on MTC by comparing each participant's after intervention (post-training) data to that obtained in retention conditions i.e., MTC data at 1-month (short-term retention), and at 3-month, 6-month (long-term retention).
3) MTC data collected during overground walking during retention conditions will determine whether any changes in MTC are translated to overground walking.

Discussion
This innovative study will evaluate the impact of augmented sensory information for improving gait function, speci cally foot-ground clearance, via visually presented biofeedback. The research ndings will contribute to the broader falls prevention initiative by demonstrating the effectiveness of toe clearance biofeedback in making walking safer. It will also con rm whether faded feedback enhances learning by demonstrating how performance is affected immediately following training and whether it can be retained in the long term. If shown to be effective, there will be opportunities for biofeedback-training applications to other gait-impaired clinical populations. The project's ndings will contribute to understanding how stroke-related changes to sensory and motor processes in uence the tripping risks of everyday walking. The protocol is, however, limited to people with chronic stroke capable of walking on a treadmill due to the continuous real-time feedback used for gait training.

Declarations
Ethics approval and consent to participate Ethics approval for the study was obtained from Austin Health, Melbourne Health and Victoria University Human Research Ethics Committees. The approving HREC is Austin Health. HREC/15/Austin/516. Written informed consent is obtained from all participants using the procedures mandated and approved by the above institutional research ethics committees. The blinded assessor (clinical physiotherapist) will obtain informed consent during the rst visit for clinical assessment. Patients are fully informed that they are free to withdraw at any time without providing a reason.   SPIRIT study-timeline. <b>T1 to T10</b> gait training sessions of (up to) 10 min treadmill walking.

Consent for publication
Feedback (FB) group given visual feedback display of minimum toe clearance (MTC), No-Feedback group given no gait-related information<b>. Gait Bio. Baseline: </b>Gait biomechanics variables for overground and treadmill walking.<b> Fade</b>: MTC information reduced progressively for FB group. <b>COPE</b>: Centre of Pressure Excursion. Gait stability assessment from plantar pressures. <b>Clinical Tests, Borg Scales and Falls Calendars</b>: see text.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.